Direction Finding for Tracking Radar Using a Simplified, Subspace Based Approach.

Dr. S. D. Hayward and Mr.  A. R. Green
DERA Malvern
St Andrews Road
Malvern Worcs WR14 1SA


Abstract Multiple coherent reflections from different parts of a radar target, or multiple reflection paths through a transmission environment can lead to errors when estimating their directions of arrival (DOA). This problem is particularly prevalent in radar tracking systems based on monopulse, a technique which compares the outputs from two conventionally for med beams to measure the target's angular offset from a known 'look' direction. These errors, commonly known as glint, arise because of the incorrect assumption that a received signal wave front emanating from a far field target will be planar and will be normal to the target's DOA. The signals from several closely spaced point reflectors within a target can combine to create singularities in the scattered field. These are small regions of space where severe destructive interference leads to signal fading and where local curvature in the wave fronts means that they are no longer normal to the target direction. Various subspace ideas have been proposed for the general resolution of multiple coherent signals using phased array receivers. Particularly relevant are MUSIC when combined with an interpolated version of spatial smoothing, and the IMP algorithm. Both methods model the target signals by a multi-dimensional subspace and attempt to resolve individual scatterers. We present a simplification of these ideas, specifically applicable to radar tracking systems, in which a subspace of very low dimensionality is used to approximate the signal from a cluster of many closely spaced scatterers. The concept of 'distance' between the signal subspace and a fixed reference subspace is used directly to measure the target position without the need for a peak search or polynomial rooting algorithm. A suitable projection based algorithm is described which is comparable in complexity to the monopulse method and this is shown to give mean DOA estimation errors which are robust against glint.

Presentation in PDF format